Leakage detection in pipeline systems using machine learning

Abstract

The reliability of pipeline systems as a criterion is of enormous significance in sustainable pipeline operation and the protection of the environment. In their basic form, conventional leak detection techniques are often slow and not sensitive enough to suit many purposes, particularly in the early detection and control of leaks in large distributed systems. In this paper, we examine the application of machine learning—One-Class Support Vector Machine (SVM)—to the existing pipeline leak detection systems. Using both COMSOL Multiphysics for simulation and MATLAB for data analysis, this work proves that machine learning is applicable to improve leakage assessment. Using detailed simulations under various operational conditions, the k coefficients of the One-Class SVM model pinpoint pressure, temperature, and velocity abnormalities that suggest leakage. The results also clearly indicate the model’s effectiveness in accurately identifying leak locations in addition to simply identifying their presence, making it a significant improvement over current approaches by increasing response speed while decreasing possible losses and threats to the environment.